JP7729467B2 - Learning Device - Google Patents

Description

The present invention relates to a learning device, a learning method, a recording medium, and an inference device.

The learning machines trained by each participant may be combined in order to achieve better performance, etc.

For example, Non-Patent Document 1 describes a technique called Gradient Boosting Forest (GBF), in which participants each create a decision tree at each step and combine the created decision trees to create a model with better performance.

Related literature includes, for example, Patent Document 1, which describes GBDT (Gradient Boosting Decision Tree) and other techniques.

Japanese Patent Application Laid-Open No. 2021-140296

Feng Wang et al., Gradient Boosting Forest: a Two-Stage Ensemble Method Enabling Federated Learning of GBDTsICONIP 2021: Neural Information Processing pp. 75-86 [Retrieved March 8, 2022], Internet: <https://link.springer.com/chapter/10.1007/978-3-030-92270-2_7>

In the case of the technology described in Non-Patent Document 1, the learning machines trained by each participant are combined. Therefore, the ultimately created learning machine will be based on, for example, the training data possessed by each participant. On the other hand, in cases where, for example, there is a bias in the distribution of the training data possessed by each participant, it may be desirable to prepare a learning machine that is more suited to a specific individual while improving performance. However, simply combining the learning machines as described in Non-Patent Document 1 makes it difficult to handle such cases.

Therefore, the object of the present invention is to provide a learning device, learning method, and recording medium that can solve the above-mentioned problems.

In order to achieve this object, a learning device according to one embodiment of the present disclosure includes:
a receiving unit that receives a learning device from another learning device;
a generation unit that generates an appropriate learning unit using the learning unit received by the receiving unit and data stored in the device itself to perform prediction appropriate for the data stored in the device itself;
The configuration is as follows:

Furthermore, a learning method according to another aspect of the present disclosure includes:
The information processing device
receiving a learner from another learner;
The received learning device and the data held by the device itself are used to generate an appropriate learning device that makes predictions suitable for the data held by the device itself.

Furthermore, a recording medium according to another aspect of the present disclosure includes:
In the information processing device,
receiving a learner from another learner;
This is a computer-readable recording medium that stores a program for realizing the process of using a received learning device and data held by the device to generate an appropriate learning device that makes predictions appropriate for the data held by the device.

Furthermore, an inference device according to another aspect of the present disclosure includes:
An inference device that infers a label for an input of a feature,
a storage device that stores the learning devices received from other learning devices and the coupling coefficients calculated using the data stored in the learning device itself so as to perform predictions suitable for the data stored in the learning device itself;
an inference unit that performs inference using the learning devices and combination coefficients stored in the storage device in response to input of feature quantities;
The configuration is as follows:

The configurations described above make it possible to improve performance while providing a learning machine that is more suited to specific individuals.

FIG. 1 is a diagram for explaining an overview of the present disclosure. FIG. 1 is a diagram illustrating a configuration example of a learning system according to a first embodiment of the present disclosure. FIG. 2 is a block diagram illustrating an example of the configuration of a learning device. FIG. 10 is a diagram illustrating an example of learning data. FIG. 10 is a diagram illustrating an example of validation data. FIG. 10 is a diagram illustrating an example of an algorithm. FIG. 10 is a diagram illustrating another example of an algorithm. 10 is a flowchart illustrating an example of the operation of the learning device. FIG. 10 is a diagram illustrating an example of a hardware configuration of a learning device according to a second embodiment of the present disclosure. 1 is a block diagram showing an example of the configuration of a learning device; FIG. 2 is a block diagram showing an example of the configuration of an inference device.

[First embodiment]
A first embodiment of the present disclosure will be described with reference to FIGS. 1 to 8. FIG. 1 is a diagram for explaining an overview of the present disclosure. FIG. 2 is a diagram illustrating an example configuration of a learning system 100. FIG. 3 is a block diagram illustrating an example configuration of a learning device 300. FIG. 4 is a diagram illustrating an example of learning data information 341. FIG. 5 is a diagram illustrating an example of validation data information 342. FIGS. 6 and 7 are diagrams illustrating examples of algorithms performed in the learning system 100. FIG. 8 is a flowchart illustrating an example operation of the learning device 300.

In a first embodiment of the present disclosure, as shown in FIG. 1, a learning system 100 is described that includes a learning device 300 that generates a combined learning device to make more appropriate predictions for data such as training data and validation data held by the learning device itself, based on learning devices received from other participants in the learning system 100 and learning devices trained on the learning device itself. As will be described later, when the learning device 300 described in this embodiment receives a decision tree that is a learning device from another participant in the learning system 100, it calculates new features by inputting the training data held by the learning device itself into the received decision tree. Then, the learning device 300 generates a decision tree that is a new learning device for the learning device itself, based on the training data and the calculated features.

The learning device 300 also calculates a combining coefficient corresponding to each decision tree using the generated decision tree, decision trees received from other participants, and pre-stored validation data. For example, the learning device 300 calculates a combining coefficient for each decision tree based on the decision tree and the validation data so as to optimize its own predictive performance for the validation data. The learning device 300 then uses the calculated combining coefficient to combine decision trees received from other learning devices 200 or the generated decision tree with past combined decision trees to generate a new combined decision tree. In other words, the learning device 300 generates a combined decision tree, which is an appropriated learner, by performing combinations using the combining coefficients.

For example, as shown in Figure 1, the learning system 100 described in this embodiment can repeat the above-mentioned processing. In other words, the learning system 100 can repeat a series of steps, such as another learning device 200 generating a decision tree, a learning device 300 generating a decision tree using the decision tree generated by the other learning device 200, calculating a combination coefficient, and combining the decision trees, multiple times until, for example, a predetermined condition is met.

The training data includes multiple features such as gender, age, height, weight, etc., as well as labels indicating whether or not the subject is ill. Features may also be called explanatory variables or attributes. Labels may also be called objective variables. Specific examples of features and labels may be other than those listed above. Validation data refers to data used for verification purposes that can be used, for example, to evaluate a learning device. Validation data, like training data, includes multiple features and labels. A decision tree refers to a model that is trained by repeatedly sorting input data using a binary tree based on conditional branching of the features until the label's explanatory performance is sufficiently good. Decision trees include regression trees that handle regression tasks and classification trees that handle classification tasks.

Furthermore, in this embodiment, a case will be described in which a decision tree is used as a learning device. However, the learning device that is the subject of the present invention is not limited to a decision tree. For example, each participant in the learning system 100 may generate a shallow neural network, a support vector machine, or the like as a learning device. Even if each participant in the learning system 100 generates a neural network, a support vector machine, or the like as a learning device, the present invention can be applied without any problems.

Figure 2 shows an example of the overall configuration of the learning system 100. Referring to Figure 2, the learning system 100 has one or more other learning devices 200 and a learning device 300. As shown in Figure 2, the other learning devices 200 and the learning device 300 are connected via a network or the like so that they can communicate with each other.

The other learning device 200 is an information processing device that generates a decision tree, which is a learner, by performing learning based on the training data held by the other learning device 200. The other learning device 200 can also transmit the generated decision tree to other other learning devices 200 or the learning device 300.

The other learning device 200 may also be configured to combine decision trees, which are learning machines received from other learning devices 200, learning devices 300, etc., using a predetermined coefficient or the like. Here, the predetermined coefficient can be determined in advance, for example, based on the amount of data held by each participant, such as the other learning devices 200 and learning devices 300, in the learning system 100. For example, the predetermined coefficient may be calculated by dividing the amount of training data held by the device itself by the sum of the amount of training data held by the other learning devices 200 and learning devices 300 in the learning system 100.

For example, as described above, the other learning device 200 generates a decision tree, which is a learner, based on the training data it has, and transmits the generated decision tree to other information processing devices within the learning system 100. The other learning device 200 can also receive decision trees from other information processing devices within the learning system 100 and combine the received decision trees using predetermined coefficients, etc. For example, the other learning device 200 may implement each of the above processes using a method such as that described in Non-Patent Document 1.

The learning device 300 is an information processing device that generates a combined decision tree, which is a combined learner, to make more appropriate predictions for data such as training data and validation data that the device holds. Figure 3 shows an example configuration of the learning device 300. Referring to Figure 3, the learning device 300 has, as its main components, an operation input unit 310, a screen display unit 320, a communication I/F unit 330, a memory unit 340, and an arithmetic processing unit 350, for example.

Note that Figure 3 illustrates an example in which the functions of the learning device 300 are realized using a single information processing device. However, the learning device 300 may be realized using multiple information processing devices, for example, by being realized on the cloud. Furthermore, the learning device 300 may not include some of the configurations exemplified above, such as not having the operation input unit 310 or the screen display unit 320, or may have configurations other than those exemplified above.

The operation input unit 310 consists of operation input devices such as a keyboard and a mouse. The operation input unit 310 detects operations performed by the operator operating the learning device 300 and outputs the operations to the calculation processing unit 350.

The screen display unit 320 consists of a screen display device such as an LCD (Liquid Crystal Display). The screen display unit 320 can display various information stored in the memory unit 340 on the screen in response to instructions from the calculation processing unit 350.

The communication I/F unit 330 consists of a data communication circuit, etc. The communication I/F unit 330 performs data communication with an external device connected via a communication line.

The memory unit 340 is a storage device such as a hard disk or memory. The memory unit 340 stores processing information and programs 345 required for various processes in the calculation processing unit 350. The programs 345 are read into the calculation processing unit 350 and executed to realize various processing units. The programs 345 are read in advance from an external device or recording medium via a data input/output function such as the communication I/F unit 330, and are stored in the memory unit 340. Examples of the main information stored in the memory unit 340 include learning data information 341, validation data information 342, learning device information 343, and coefficient information 344.

The learning data information 341 includes learning data used when training the decision tree, which is the learning device. For example, the learning data information 341 is acquired in advance using a method such as obtaining it from an external device via the communication I/F unit 330 or inputting it using the operation input unit 310, and is stored in the memory unit 340.

Fig. 4 shows an example of the training data information 341. Referring to Fig. 4, in the training data information 341, a plurality of feature quantities are associated with labels. For example, in the example shown in Fig. 4, feature quantities ( _x1 , _x2 , ..., _xd ) are associated with a label _y1 . As shown in Fig. 4, the training data information 341 may include a plurality of training data.

Validation data information 342 includes validation data, which is data used when verifying the performance of a decision tree, etc. For example, validation data information 342 is acquired in advance using methods such as obtaining it from an external device via communication I/F unit 330 or inputting it using operation input unit 310, and is stored in memory unit 340.

Fig. 5 shows an example of the validation data information 342. Referring to Fig. 5, in the validation data information 342, similar to the training data information 341, a plurality of feature quantities are associated with labels. For example, in the example shown in Fig. 5, feature quantities (x ₁₁ , x ₁₂ , ..., x _1d ) are associated with a label y _10. As shown in Fig. 5, the validation data information 342 may include a plurality of validation data.

The learner information 343 includes information indicating decision trees received from other learning devices 200 and combined decision trees combined by the combining unit 355 described below. For example, the learner information 343 associates decision trees with identification information indicating the source of the decision tree. The learner information 343 may include decision trees and combined decision trees for each step. For example, the learner information 343 is updated when the receiving unit 351 described below receives a decision tree from another learning device 200, when the combining unit 355 generates a combined decision tree, etc.

The coefficient information 344 includes a combination coefficient corresponding to each decision tree, such as a decision tree received from another learning device 200 or a decision tree generated by the learning unit 353 described below. For example, the coefficient information 344 associates the identification information of the decision tree with the combination coefficient. The coefficient information 344 may include a combination coefficient for each step and each decision tree. For example, the coefficient information 344 is updated in response to the calculation of the combination coefficient by the coefficient calculation unit 354 described below.

The arithmetic processing unit 350 has an arithmetic device such as a CPU (Central Processing Unit) and its peripheral circuits. The arithmetic processing unit 350 reads and executes the program 345 from the storage unit 340, thereby enabling the above hardware and the program 345 to work together to realize various processing functions. The main processing units realized by the arithmetic processing unit 350 include, for example, a receiving unit 351, a feature addition calculation unit 352, a learning unit 353, a coefficient calculation unit 354, a combining unit 355, an inference unit 356, and an output unit 357.

The receiving unit 351 receives a decision tree, which is a learning device, from another learning device 200. For example, the receiving unit 351 can receive a learning device from each other learning device 200 included in the learning system 100 at each step. The receiving unit 351 also stores the received decision tree in the memory unit 340 as learning device information 343.

The receiving unit 351 may receive, for example, information indicating the difference between the decision tree in the previous step from another learning device 200. In this case, the receiving unit 351 may be configured to update the corresponding decision tree based on the received information indicating the difference.

The feature addition calculation unit 352 calculates additional learning data based on the decision tree received by the receiving unit 351 and the learning data included in the learning data information 341. For example, the feature addition calculation unit 352 obtains output from the learning device by inputting each piece of learning data included in the learning data information 341 into the decision tree received by the receiving unit 351. The feature addition calculation unit 352 can acquire the output as an additional feature.

For example, suppose the learning data information 341 includes learning data ( _xi , _yi ) including a feature _xi and a label _yi (i may be arbitrary). Also suppose that decision trees _f1 (), _f2 (), ... are received from the other learning device 200. In this case, the feature addition calculation unit 352 inputs the feature xi to each decision tree to calculate additional features _f1 ( _xi ), _f2 ( _xi ), .... As a result, the learning data to be learned by the learning unit 353 (described later) becomes ( _xi , _f1 ( _xi ), _f2 ( _xi ), ..., _yi ).

For example, the feature addition calculation unit 352 can perform the above-described processing for each decision tree and for each piece of training data. For example, the feature addition calculation unit 352 may perform the above-described processing for each decision tree extracted using any method.

The learning unit 353 generates a decision tree, which is a learning device, by performing learning based on the features calculated by the feature addition calculation unit 352 and the learning data indicated by the learning data information 341. The learning unit 353 also stores the generated decision tree in the memory unit 340 as learning device information 343.

For example, as described above, additional features are calculated by the feature addition calculation unit 352. Therefore, the learning unit 353 generates a decision tree by performing machine learning using learning data including the additional features, such as (x _i , f ₁ (x _i ), f ₂ (x _i ), ..., y _i ).

The learning unit 353 may perform machine learning by adding the additional features calculated by the feature addition calculation unit 352 directly to the training data as described above, or may be configured to perform machine learning by adding, for example, the results of linearly combining the additional features calculated by the feature addition calculation unit 352 to the training data. The learning unit 353 may perform machine learning by adding both the additional features calculated by the feature addition calculation unit 352 and the results of linearly combining the additional features to the training data.

The coefficient calculation unit 354 calculates a combining coefficient for each decision tree using the validation data indicated by the validation data information 342. For example, the coefficient calculation unit 354 calculates the combining coefficient so as to optimize the predictive performance for the validation data indicated by the validation data information 342. The coefficient calculation unit 354 can calculate a combining coefficient for each decision tree received by the receiving unit 351 or each decision tree generated by the learning unit 353. The coefficient calculation unit 354 also stores the calculated combining coefficient in the memory unit 340 as coefficient information 344.

For example, it is assumed that the validation data information 342 includes validation data (x _1i , y _1i ) including feature x _{1i and} label y _1i . It is also assumed that decision trees f ₁₁ (), f ₁₂ (), ... are received or generated by another learning device 200 or the learning unit 353. In this case, the coefficient calculation unit 354 first obtains an output by inputting the validation data to each decision tree. For example, the coefficient calculation unit 354 obtains an output u _i by inputting the validation data (x _1i , y _1i ) to the decision tree f ₁₁ (). The coefficient calculation unit 354 also obtains an output v _i by inputting the validation data (x _1i , y _1i ) to the decision tree f ₁₂ (). The coefficient calculation unit 354 then calculates a coupling coefficient for each decision tree using (u _i , v _i , y _1i ). For example, the coefficient calculation unit 354 may calculate the coupling coefficient by performing linear regression, etc. For example, the coefficient calculation unit 354 may determine the coupling coefficient corresponding to the decision tree f11() by performing linear regression using the validation data (x _1i , y _1i ) and the output u _i .

Specifically, for example, the coefficient calculation unit 354 can determine the coefficient ai corresponding to each decision tree by performing linear regression on Equation 1 using the validation data ( _x1i , _{y1i )} and _ui and _vi .

The coefficient calculation unit 354 may calculate the coupling coefficients using the entire validation data, or may calculate the coupling coefficients using only a portion of the validation data. For example, by referencing model information about the decision tree generated by the learning unit 353, such as the model structure and branching conditions, it is possible to identify the leaf nodes into which each validation data falls. Therefore, the coefficient calculation unit 354 may calculate the coupling coefficients for each leaf node, for example, by performing linear regression using the validation data for each leaf node. Even when calculating the coupling coefficients for each leaf node, a coupled decision tree can be generated in the same way as in the above case by coupling each leaf node. When the coefficient calculation unit 354 calculates the coupling coefficients using the entire validation data, it can also be said that the coupling coefficients for the entire decision tree are calculated.

The coefficient calculation unit 354 may also calculate the coupling coefficient using a method other than the above examples. For example, the coefficient calculation unit 354 may calculate the coupling coefficient using the learning data indicated by the learning data information 341 instead of the validation data. However, from the perspective of suppressing excessive bias, it is more desirable to calculate the coupling coefficient using the validation data rather than the learning data. The coefficient calculation unit 354 may also calculate the coupling coefficient using any other method.

The combining unit 355 generates a new combined decision tree by combining the decision tree received by the receiving unit 351 or the decision tree generated by the learning unit 353 with the combined decision tree one step earlier included in the learning device information 343, using the combining coefficient calculated by the coefficient calculation unit 354. The combining unit 355 also stores the newly generated combined decision tree in the memory unit 340 as learning device information 343.

For example, the combining unit 355 performs combining using a combining coefficient by solving the following equation (2).
Here, f ^(t-1) indicates the combined decision tree one step before, and f ^(t) indicates the newly generated combined decision tree. Also, a _k indicates a combining coefficient, and f _k indicates the decision tree received by the receiving unit 351 or the decision tree generated by the learning unit 353. k is a value corresponding to, for example, the number of other learning devices 200 and learning devices 300 included in the learning system 100.

The inference unit 356 performs inference using the combined decision tree. For example, the inference unit 356 can perform inference using the latest combined decision tree.

As shown in Equation 3, the latest combined decision tree f _final () includes decision trees, combining coefficients, etc. received or generated in past steps. Therefore, the inference unit 356 can also perform inference using decision trees, combining coefficients, etc. generated in past steps.
As shown in Equation 4, ft() is a term added to the model at the t-th step, and is the linear sum of the decision trees _fk () created by each participant.
Note that Equation 5 indicates the coefficient of the decision tree created by the kth participant at the tth step.

The output unit 357 outputs decision trees and combined decision trees, and outputs the inference results of the inference unit 356.

For example, the output unit 357 can transmit the decision tree generated by the learning unit 353 or the combined decision tree generated by the combining unit 355 to an external device such as another learning device 200 via the communication I/F unit 330. The output unit 357 may output at any timing, such as after processing by the learning unit 353 or the combining unit 355.

In addition, the output unit 357 can display the results of the inference by the inference unit 356 on the screen display unit 320 or transmit them to an external device via the communication I/F unit 330.

The above is an example configuration of the learning device 300. Note that FIG. 6 shows an example algorithm of the learning system 100 when dealing with a regression task, as described in Non-Patent Document 1. The example shown in FIG. 6 illustrates a case where participants such as other learning devices 200 and learning devices 300 are included in the K-class learning system 100. As illustrated in FIG. 6 , the learning system 100 combines the decision trees f _k () generated by each participant using a combination coefficient a. As described above, the combination coefficient a is calculated so as to optimize the prediction performance for the validation data. Therefore, by performing combination using the combination coefficient a, it is possible to improve performance and generate a combined decision tree in a form that is more suitable for the learning device 300 that has validation data.

Note that the example shown in Figure 6 illustrates an example algorithm for the learning system 100 when dealing with a regression task; however, even when dealing with a classification task instead of a regression task, combining is performed using the combining coefficient a, just as in the case of a regression task. Therefore, even when dealing with a classification task, combining using the combining coefficient a, just as in the case of a regression task, can improve performance while generating a combined decision tree that is more suitable for a learning device 300 that has validation data. For example, Non-Patent Document 1 describes an example algorithm for dealing with a classification task as shown in Figure 7. Referring to Figure 7, it can be seen that even when dealing with a classification task, combining is performed using the combining coefficient a, just as in the case of a regression task.

Next, an example of the operation of the learning device 300 will be described with reference to Figure 8. Figure 8 is a flowchart showing an example of the operation of the learning device 300. Referring to Figure 8, the receiving unit 351 receives a decision tree, which is a learning device, from another learning device 200 (step S101).

The feature addition calculation unit 352 calculates additional learning data based on the decision tree received by the receiving unit 351 and the learning data included in the learning data information 341 (step S102). For example, the feature addition calculation unit 352 obtains output from the learning device by inputting each piece of learning data included in the learning data information 341 into the decision tree received by the receiving unit 351. The feature addition calculation unit 352 can acquire the output as an additional feature.

The learning unit 353 generates a decision tree, which is a learning device, by performing learning based on the features calculated by the feature addition calculation unit 352 and the learning data indicated by the learning data information 341 (step S103). In other words, the learning unit 353 generates a decision tree by performing learning by adding the features calculated by the feature addition calculation unit 352 to the learning data indicated by the learning data information 341.

The coefficient calculation unit 354 calculates a combining coefficient for each decision tree using the validation data indicated by the validation data information 342 (step S104). For example, the coefficient calculation unit 354 calculates the combining coefficient so as to optimize the predictive performance for the validation data indicated by the validation data information 342. The coefficient calculation unit 354 can calculate a combining coefficient for each decision tree received by the receiving unit 351 or each decision tree generated by the learning unit 353.

The combining unit 355 uses the combining coefficient calculated by the coefficient calculation unit 354 to combine the decision tree received by the receiving unit 351 or the decision tree generated by the learning unit 353 with the combined decision tree one step earlier included in the learning device information 343 to generate a new combined decision tree (step S105).

The above is an example of the operation of the learning device 300.

In this way, the learning device 300 has a feature addition calculation unit 352 and a learning unit 353. With this configuration, the learning unit 353 can generate a decision tree that is a learning device by performing learning using training data to which features calculated by the feature addition calculation unit 352 have been added. As a result, a decision tree can be generated that incorporates the results of learning by other learning devices 200. This makes it possible to improve performance while generating a decision tree that is a learning device that is more suitable for the data stored in the device itself.

The learning device 300 also has a coefficient calculation unit 354 and a combination unit 355. With this configuration, the combination unit 355 can calculate each decision tree using the combination coefficients calculated by the coefficient calculation unit 354. As described above, the combination coefficients are calculated so as to optimize the predictive performance for the validation data. Therefore, by combining each decision tree using the above combination coefficients, it is possible to improve performance and generate a decision tree that is a learning device more suitable for the learning device 300 that has validation data.

In this embodiment, an example is given of the case where the learning device 300 has both the feature addition calculation unit 352 and the coefficient calculation unit 354. However, the learning device 300 may have only one of the feature addition calculation unit 352 and the coefficient calculation unit 354.

For example, if the learning device 300 does not have the feature addition calculation unit 352, the learning unit 353 generates a decision tree by performing learning based on the learning data included in the learning data information 341. Even in such a case, as described above, the combination coefficients are calculated to optimize the predictive performance for the validation data, and by combining each decision tree using the combination coefficients, it is possible to generate a decision tree that is a learning machine more suitable for the learning device 300 that has the validation data while improving performance.

Furthermore, for example, if the learning device 300 does not have a coefficient calculation unit 354, the combining unit 355 combines a decision tree generated by incorporating the results of learning by other learning devices 200 with the combined decision tree from one step earlier. As a result, it is possible to generate a decision tree that is a learner that is more suitable for the data held by the device itself while improving performance. Note that if the learning device 300 does not have a coefficient calculation unit 354, the combining unit 355 may combine a decision tree generated without using a combining coefficient with the combined decision tree from one step earlier.

As mentioned above, in this embodiment, a decision tree is used as the learning device. However, the learning device that is the subject of the present invention is not limited to a decision tree. For example, the learning device may be a shallow neural network or a support vector machine. Of course, the learning device that is ultimately generated may also be one that corresponds to each of the above learning devices. For example, when a decision tree is used as the learning device, the ultimately generated combined decision tree is a GBDT (Gradient Boosting Decision Tree) model. Furthermore, when a neural network is used as the learning device, the ultimately generated model is a Gradient Boosting Neural Network.

Second Embodiment
Next, a second embodiment of the present disclosure will be described with reference to Fig. 9 to Fig. 11. Fig. 9 is a diagram illustrating an example of the hardware configuration of a learning device 400. Fig. 10 is a block diagram illustrating an example of the configuration of the learning device 400. Fig. 11 is a block diagram illustrating an example of the configuration of an inference device 500.

In the second embodiment of the present disclosure, a configuration example of a learning device 400 will be described, which is an information processing device that receives learning devices from other devices and combines the learning devices to perform optimal predictions on its own data. Fig. 9 shows an example of the hardware configuration of the learning device 400. Referring to Fig. 9, the learning device 400 has, as an example, the following hardware configuration.
・CPU (Central Processing Unit) 401 (arithmetic unit)
ROM (Read Only Memory) 402 (storage device)
RAM (Random Access Memory) 403 (storage device)
Program group 404 loaded into RAM 403
A storage device 405 for storing the program group 404
A drive device 406 that reads and writes data from and to a recording medium 410 outside the information processing device
A communication interface 407 that connects to a communication network 411 outside the information processing device
Input/output interface 408 for inputting and outputting data
A bus 409 that connects each component

The learning device 400 can realize the functions of the receiving unit 421 and generating unit 422 shown in FIG. 10 by having the CPU 401 acquire and execute the program group 404. The program group 404 is stored in advance in, for example, the storage device 405 or ROM 402, and is loaded into the RAM 403 or the like by the CPU 401 for execution as needed. The program group 404 may be supplied to the CPU 401 via the communication network 411, or may be stored in advance on the recording medium 410, with the drive device 406 reading out the programs and supplying them to the CPU 401.

Note that Figure 9 shows an example hardware configuration of the learning device 400. The hardware configuration of the learning device 400 is not limited to the above-described case. For example, the learning device 400 may be configured with only a part of the above-described configuration, such as not including the drive device 406.

The receiving unit 421 receives a learning device from another learning device.

The generation unit 422 uses the learner received by the receiving unit 421 and the data held by its own device to generate an appropriate learning device that makes predictions appropriate for the data held by its own device. For example, the generation unit 422 generates an appropriate learning device that makes predictions appropriate for the data held by its own device by combining the learners received by the receiving unit 421 using a combination coefficient calculated using the data held by its own device. Alternatively, the generation unit 422 generates an appropriate learning device by generating a learning device using learning data to which additional features calculated using the learner received by the receiving unit have been added. For example, the generation unit 422 can generate an appropriate learning device by performing any of the methods exemplified above, or a combination thereof.

As such, the learning device 400 has a receiving unit 421 and a generating unit 422. With this configuration, the generating unit 422 can use the learning device received by the receiving unit 421 and the data held by the device itself to generate an appropriate learning device that makes predictions appropriate for the data held by the device itself. As a result, it is possible to generate a learning device that is more appropriate for the data held by the device itself while improving performance.

The above-mentioned learning device 400 can be realized by incorporating a predetermined program into an information processing device such as the learning device 400. Specifically, another form of the program of the present invention is a program for realizing a process in which an information processing device such as the learning device 400 receives a learning device from another learning device, and uses the received learning device and data held by the device to generate an appropriate learning device that makes predictions appropriate for the data held by the device.

In addition, the learning method executed by an information processing device such as the above-mentioned learning device 400 is a method in which an information processing device such as the learning device 400 receives a learning device from another learning device, and uses the received learning device and data held by the device to generate an appropriate learning device that makes predictions appropriate to the data held by the device.

An invention of a program, a computer-readable recording medium having a program recorded thereon, or a learning method having the above-described configuration can achieve the same functions and effects as the above-described learning device 400, and therefore can achieve the above-described objective of the present disclosure.

Furthermore, the objectives of the present disclosure can be similarly achieved with an inference device 500 that performs inference using an appropriated learning device generated by the above-mentioned learning device 400 or the like. For example, as shown in FIG. 11, the inference device 500 has a storage device 521 that stores a learning device received from another learning device and a combination coefficient calculated using data held by the device itself to make a prediction appropriate to the data held by the device itself, and an inference unit 522 that performs inference using the learning device and combination coefficient stored in the storage device in response to input of feature quantities. Note that the hardware configuration of the inference device 500 may be the same as that of the learning device 400.

<Additional Notes>
A part or all of the above-described embodiments can be described as follows: The learning device and the like according to the present invention will be outlined below. However, the present invention is not limited to the following configuration.

(Appendix 1)
a receiving unit that receives a learning device from another learning device;
a generation unit that generates an appropriate learning unit using the learning unit received by the receiving unit and data stored in the device itself to perform prediction appropriate for the data stored in the device itself;
A learning device having:
(Appendix 2)
2. The learning device according to claim 1,
a calculation unit that calculates a coupling coefficient using data stored in the device itself so as to perform a prediction suitable for the data;
The generation unit generates the appropriate learning device by combining the learning devices received by the receiving unit using the combination coefficient calculated by the calculation unit so as to make predictions appropriate for data held by the device itself.
(Appendix 3)
3. The learning device according to claim 2,
The calculation unit performs linear regression using the results of inputting data held by the device into a learning device, thereby calculating a coupling coefficient so as to make a prediction suitable for the data.
(Appendix 4)
4. The learning device according to claim 2, wherein:
The calculation unit calculates the coupling coefficient using validation data, which is data for validation.
(Appendix 5)
1. The learning device according to claim 1, wherein:
The calculation unit identifies data that falls at each leaf node in a decision tree that is a learning device, and calculates the coupling coefficients using the data for each leaf node.
(Appendix 6)
10. The learning device according to claim 1, wherein:
a feature calculation unit that calculates additional features using the learning device and training data received by the receiving unit;
a learning unit that generates a learning device by performing learning by adding the feature calculated by the feature calculation unit to the training data;
and
The generation unit generates the appropriated learning device using the learning device generated by the learning unit.
(Appendix 7)
7. The learning device according to claim 6,
a calculation unit that calculates a coupling coefficient corresponding to the learning device received by the receiving unit using data stored in the device itself;
The generation unit generates the appropriated learning device by combining the learning device received by the receiving unit and the learning device generated by the learning unit using the combination coefficient calculated by the calculation unit.
(Appendix 8)
The information processing device
receiving a learner from another learner;
A learning method that uses a received learning device and data held by the device itself to generate an appropriate learning device that makes predictions appropriate for the data held by the device itself.
(Appendix 9)
In the information processing device,
receiving a learner from another learner;
A computer-readable recording medium that stores a program for realizing a process of generating an appropriate learning device using the received learning device and data held by the device itself to make predictions appropriate for the data held by the device itself.
(Appendix 10)
An inference device that infers a label for an input of a feature,
a storage device that stores the learning devices received from other learning devices and the coupling coefficients calculated using the data stored in the learning device itself so as to perform predictions suitable for the data stored in the learning device itself;
an inference unit that performs inference using the learning devices and combination coefficients stored in the storage device in response to input of feature quantities;
An inference device having:

The present invention has been described above with reference to the above-mentioned embodiments, but the present invention is not limited to the above-mentioned embodiments. Various modifications that would be understood by a person skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

100 Learning system 200 Other learning device 300 Learning device 310 Operation input unit 320 Screen display unit 330 Communication I/F unit 340 Memory unit 341 Learning data information 342 Validation data information 343 Learning device information 344 Coefficient information 345 Program 350 Arithmetic processing unit 351 Receiving unit 352 Feature addition calculation unit 353 Learning unit 354 Coefficient calculation unit 355 Combining unit 356 Inference unit 357 Output unit 400 Learning device 401 CPU
402 ROM
403 RAM
404 Program group 405 Storage device 406 Drive device 407 Communication interface 408 Input/output interface 409 Bus 410 Recording medium 411 Communication network 421 Receiving unit 422 Generating unit 500 Inference device 521 Storage device 522 Inference unit

Claims (8)

a receiving unit that receives a learning device from another learning device;
a generation unit that generates an appropriate learning unit using the learning unit received by the receiving unit and data stored in the device itself to perform prediction appropriate for the data stored in the device itself;
a calculation unit that calculates a coupling coefficient using data stored in the device itself so as to perform a prediction suitable for the data ;
and
the generation unit combines the learning devices received by the reception unit using the combination coefficient calculated by the calculation unit to generate the appropriate learning device so as to perform prediction appropriate for data stored in the device itself;
The calculation unit identifies data that falls at each leaf node in a decision tree that is a learning device, and calculates the coupling coefficient using the data for each leaf node.
Learning device. The learning device according to claim 1 ,
The calculation unit performs linear regression using the results of inputting data held by the device into a learning device, thereby calculating a coupling coefficient so as to make a prediction suitable for the data. 3. The learning device according to claim 1 or 2 ,
The calculation unit calculates the coupling coefficient using validation data, which is data for validation. The learning device according to any one of claims 1 to 3 ,
a feature calculation unit that calculates additional features using the learning device and training data received by the receiving unit;
a learning unit that generates a learning device by performing learning by adding the feature calculated by the feature calculation unit to the training data;
and
The generation unit generates the appropriated learning device using the learning device generated by the learning unit. The learning device according to claim 4 ,
a calculation unit that calculates a coupling coefficient corresponding to the learning device received by the receiving unit using data stored in the device itself;
The generation unit generates the appropriated learning device by combining the learning device received by the receiving unit and the learning device generated by the learning unit using the combination coefficient calculated by the calculation unit. The information processing device
receiving a learner from another learner;
Using the received learning device and the data stored in the device, generate an appropriate learning device to perform predictions appropriate for the data stored in the device;
When generating the appropriate learning device, a combination coefficient is calculated using data held by the device itself so as to make a prediction suitable for the data, and the calculated combination coefficient is used to combine the received learning devices, thereby generating the appropriate learning device so as to make a prediction suitable for the data held by the device itself;
When calculating the coupling coefficients, data that falls at each leaf node in the decision tree, which is a learning device, is identified, and the coupling coefficients are calculated using the data for each leaf node.
How to learn. In the information processing device,
receiving a learner from another learner;
By using the received learning device and the data stored in the device, a process of generating an appropriate learning device to perform predictions appropriate for the data stored in the device is realized;
When generating the appropriate learning device, a combination coefficient is calculated using data held by the device itself so as to make a prediction suitable for the data, and the calculated combination coefficient is used to combine the received learning devices, thereby generating the appropriate learning device so as to make a prediction suitable for the data held by the device itself;
When calculating the coupling coefficients, data that falls at each leaf node in the decision tree, which is a learning device, is identified, and the coupling coefficients are calculated using the data for each leaf node.
program. a receiving unit that receives a learning device from another learning device;
a generation unit that generates an appropriate learning unit using the learning unit received by the receiving unit and data stored in the device itself to perform prediction appropriate for the data stored in the device itself;
a feature calculation unit that calculates additional features using the learning device and training data received by the receiving unit;
a learning unit that generates a learning device by performing learning by adding the feature calculated by the feature calculation unit to the training data;
and
The generation unit generates the appropriate learning device using the learning device generated by the learning unit.
Learning device.